Comparison between different features and predictive models to find proteins similar to C-peptide

Abstract

The C-peptide is a short protein that connects the A and B chains’ in the insulin prohrmone. Since its de-scription in 1967, it has been historically thought to be an inert and biologically non-active peptide. That is, no physiological roles or functions were attributed to it other than connecting A and B chains’ and proper folding of mature insulin. An increasing body of experimental evidence has challenged this view and purports the notion that C-peptide is bioactive, evidenced by observed signaling characteristics from in vitro experimental studies. The most pronounced is the ameliorated effect it has on diabetes induced renal and nerve dysfunction. Accord-ingly, the past decade has witnessed a renewal in C-peptide research following these findings aimed at providing a complete physiological characterization of the peptide. The most prominent of these include finding more functions and determining whether a receptor exits or not. In this research we provide the first attempt to computationally address thisendeavor. We performed an investigative study of C-peptide that spanned 75 organisms, to determine intrinsic features that can aide in detecting similar proteins. Features were extracted and 5 different encodings were used in unison with predictive models (Support vector machines, na"ive Bayes and others). A comparative study between these models was performed to determine the most suitable features that should be highly considered for the computational characterization of C-peptide.